Share your thoughts, 1 month free Claude Pro on usSee more

General Knowledge on MMLU-Pro (maj@2, maj@4, maj@8)

35.88maj@4 Accuracy

DLE (ε-sampling)-PROBFIRST

Updated 1mo ago

Evaluation Results

Method	Links
DLE (ε-sampling)-PROBFIRST 2026.04		35.88	33.18	37.53
DLE (top-p+top-k) 2026.04		35.73	33.22	37.28
DLE (min-p) 2026.04		35.71	33.23	36.91
DLE (ε-sampling)-DIVFIRST 2026.04		35.65	33.19	37.23
DLE (ε-sampling)-RANDBRANCH 2026.04		35.6	33.22	37.49
Self-consistency (ε-sampling) 2026.04		35.37	30.98	37.36
Self-consistency (min-p) 2026.04		34.73	30.68	36.08
Self-consistency (top-p+top-k) 2026.04		33.64	29.74	36.17
Self-consistency 2026.04		29.31	24.88	33.04
DLE (min-p) 2026.04		17.4	16.45	17.89
DLE (ε-sampling) 2026.04		17.37	16.64	17.97
DLE (top-p&top-k) 2026.04		17.3	16.36	18.19
Self-consistency (top-p&top-k) 2026.04		16.66	14.96	17.35
Self-consistency 2026.04		16.43	16.15	17.84
Self-consistency (min-p) 2026.04		16.36	14.99	17.26
Diverse beam search 2026.04		16.2	-	16.58
Self-consistency (ε-sampling) 2026.04		16.07	14.89	17.2
Self-certainty 2026.04		15.92	14.51	14.68
DeepConf 2026.04		15.58	14.41	16.76
Beam search 2026.04		15.29	15.49	15.01
Self-consistency 2026.04		15.15	13.61	16.24